Analysis of Su1Vx7¶
Analysis / figures of results from h_DsRnn_Prep_spa.ipynb
Summarizing mult. RNNs / data sets based on Pandas ExTb and ProcTb.
PREC: h_DsRnnAna_Pd.iphnbjupyter nbconvert --TemplateExporter.exclude_raw=True --TemplateExporter.exclude_input=True --to html --output tmp h_DsRnnAnaPd_Su1Vx7.ipynb
jupyter nbconvert h_DsRnnAnaPd_Su1Vx7.ipynb \
--to html \
--output tmp \
--TemplateExporter.exclude_raw=True \
--TemplateExporter.exclude_input=True \
--TagRemovePreprocessor.remove_cell_tags="{'rmCell'}" \
--TagRemovePreprocessor.remove_all_outputs_tags="{'rmOut'}" \
--RegexRemovePreprocessor.patterns="[r'(?ms).*DeprecationWarning.*']" \
--RegexRemovePreprocessor.patterns="[r'(?ms)Using TensorFlow.*']" \Add this cell metadata (JSON) e.g: { "tags": ["rmOut"]}
Imports¶
ProcTb: Load model parameter sets¶
Out[3]:
Out[4]:
[ Load ExTb - need this? ]¶
Load models / ProcDirs¶
Out[6]:
Convert Pars to Pandas (ProcTb or SecTb)¶
Out[9]:
Check training success¶
Out[13]:
Make Xarray from Pars (~SecTb)¶
Out[14]:
Out[19]:
Compute Metrics¶
Out[29]:
Boxplot Metric stats¶
Table of median per section¶
Out[33]:
Which Su data best predicts it's own left-out scans?¶
Out[35]:
Which experiments are best predicted by each subject's data?¶
Out[36]:
Plot best signals¶
Top sections for 3 metrics¶
Top 12 sections for Pcorr metric¶
Top 3 Pcorr for each Subject¶
Carpet plots?¶
Out[40]:
Out[112]:
Scratch¶
Looks like Su 65 and 6 (or 78?) show the best results on these metrics.